It takes three crucial things to build a cutting-edge, and reliable, artificial intelligence (AI) system.
First, you need a corpus of training data. Second, you need to algorithmically weight the inputs from that training data so that the AI system’s probability engine can generate accurate responses. Third, and perhaps most importantly, you need the compute power and resources to effectively implement and scale the prior two steps.
Meta CEO Mark Zuckerberg believes his company has the winning combination of all three to capture the lead in the tech sector’s ongoing AI arms race.
Fresh off of the stock market’s biggest ever single-session market value jump (a $197 billion boom), there’s a chance he may be right.
At the very least, “AI” are the two most important letters to watch in Meta’s “Year of Efficiency,” and Meta’s approach to AI — which differs from its peer competitors Google and Microsoft — has lessons for other players looking to compete in the space.
“If we succeed, everyone who uses our services will have a world-class AI assistant to help get things done. Every creator will have an AI that their community can engage with. Every business will have an AI that their customers can interact with to buy goods and get support, and every developer will have a state-of-the-art open-source model to build with,” Zuckerberg told investors on Meta’s Thursday (Feb. 1) full year 2023 earnings call.
The company plans to spend up to $37 billion a year “playing to win.”
“The number of people daily using Facebook, Instagram and WhatsApp is the highest it’s ever been,” Zuckerberg explained to investors, noting that across the Meta Platforms suite of products, monthly active users, as a metric, was 3.7 billion at the end of the year.
But how does this relate to Meta’s AI strategy?
For one, as Zuckerberg told investors as he unpacked his company’s AI strategy, “on Facebook and Instagram there are hundreds of billions of publicly shared images and tens of billions of public videos, which we estimate is greater than the Common Crawl dataset, and people share large numbers of public text posts in comments across our services as well.”
Importantly, those posts aren’t static pieces of content — they are in effect already fine-tuned and weighted by things like “likes” and “views” and other engagement metrics native to the social media experience. Metrics which Meta can leverage to continuously improve its products.
After all, establishing robust feedback loops with a large user base is crucial for refining AI systems, addressing user concerns, and ensuring continuous improvement. And Meta owns this content, protecting it from copyright lawsuits like the those faced by OpenAI.
Feedback loops allow an AI system to continuously learn and adapt based on user interactions. By collecting input from millions — and billions — of users, the system can identify patterns, correct errors, and refine its algorithms over time. This iterative process enables rapid improvements in the AI system’s performance and capabilities.
This approach not only enhances the performance and reliability of AI services but also fosters a stronger connection between the company and its user community, driving differentiation in a competitive landscape.
User behavior and preferences evolve over time. Regular feedback ensures that an AI system remains adaptable and responsive to changing trends, allowing the company to stay ahead of the curve and meet emerging user needs effectively.
While the jury is out on whether Meta’s user data is more valuable for AI training purposes when compared to the massive corpus of data used by other companies — for example, Google has its titular search engine to draw on, as well as YouTube and its G-suite of office products, while Microsoft and OpenAI are forging as many content partnerships as they can — Meta’s open-source approach to AI development provides a potential growth engine.
That’s because, as Google, Microsoft and OpenAI, and others push forward developing closed-source AI systems which can only progress as fast as the engineering talent at those firms are able to prod them, Meta is taking a different tact and moving forward with an open-source approach, allowing it to benefit from the external developers building off of its foundational AI models.
Of course, every game plan is perfect in a vacuum. As the AI arms race plays out, the strategies each big tech giant acts out will be key to watch — and certain to impact each other.